Kyrune Mini 1.0
Free uncensored roleplay API for JanitorAI, SillyTavern and Chub
Kyrune Mini is the Roleplay API's free model, uncensored on kyrune.red for adults 18 and over. Free account required. 100 requests a day, reset at 00:00 UTC. One reply at a time on Free. 28,672 tokens of context, shared by the prompt and the reply.
JanitorAI, API Settings, Proxy
Model Name
kyrune-mini-1.0-freeProxy URL
https://kyrune.red/api/v1/chat/completions API Key
kammii-Your key, from the API dashboard- $0
- Kyrune Mini on every plan, Free included
- 100 a day
- Its own requests, apart from your plan's
- 28,672 tokens
- Context, shared by the prompt and the reply
- 00:00 UTC
- When the 10 a day on other models reset
When you want more than Kyrune Mini
Plus has unlimited requests on every roleplay model and 64K of context for $12.99 a month. Your key and your setup stay the same.
What is free
Kyrune Mini 1.0 is Kyrune's own roleplay model and the Roleplay API's free model. It is free on every plan, the Free plan included, with requests of its own that never count toward your plan's daily requests: 100 a day on Free and 300 a day on paid plans.
The API speaks OpenAI chat completions, so JanitorAI, SillyTavern and Chub can use it with a URL, a model id and your key. The plan's other limits still apply: on Free, one reply at a time.
| On the Free plan | Kyrune Mini | Other models |
|---|---|---|
| Requests a day | 100, reset at 00:00 UTC | 10, reset at 00:00 UTC |
| Replies at once | 1 on Free | 1 on Free |
| Prompt size | Up to 28,672 tokens, less the reply | Up to 8,000 tokens |
| Model context | 28,672 tokens | Per model, in the table below |
| Price | $0 | $0 up to 10 a day |
What uncensored means here
kyrune.red is for adults 18 and over. Kyrune adds no content filter to Kyrune Mini's replies, and sexual content between adults is allowed.
Some things are never allowed, through the API as everywhere on Kyrune: sexual content involving a minor or anyone who looks like one, sexual content without consent, and sexual content showing a real person without their agreement. A breach of the rule on minors ends the account. The rest is in the Content Policy.
Set up JanitorAI
- Open janitorai.com, open any chat, choose the three lines menu at the top, then API Settings, then Proxy.
- Add a configuration and name it. It has three fields: Model Name, Proxy URL and API Key.
- Fill them in with the three values on this page, save, and send a message.
Model Name
kyrune-mini-1.0-freeProxy URL
https://kyrune.red/api/v1/chat/completions API Key
kammii-Your key, from the API dashboard
Connect Kyrune to JanitorAI1:30
- 0:10Open the API settings
- 0:21Provider: Proxy, then New
- 0:32Copy the address on Kyrune
- 0:42Paste the address
- 0:47Create and copy a key
- 0:57Paste the key, add a model
- 1:11Send a message
Where the video adds a model, type kyrune-mini-1.0-free.
Set up SillyTavern and Chub
SillyTavern
Choose the plug icon, set API to Chat Completion, then set Chat Completion Source to Custom (OpenAI-compatible). Paste the URL and your key, choose Connect, and pick kyrune-mini-1.0-free from the model list.
https://kyrune.red/api/v1 Do not add /chat/completions to this URL.
Chub
Open any chat, choose the three lines menu at the top, then Secrets. Choose OpenAI as the API type and turn on Reverse Proxy. Paste the URL and your key, and set the model name to kyrune-mini-1.0-free.
https://kyrune.red/api/v1 JanitorLLM down?
What free model can I use when JanitorLLM is down?
Kyrune Mini, model id kyrune-mini-1.0-free, through JanitorAI's Proxy setting with the URL https://kyrune.red/api/v1/chat/completions and a free API key. It helps when JanitorLLM itself is failing or slow; it does not fix a wrong URL, key or model id.
JanitorLLM is JanitorAI's own model. When it is failing or slow, a chat can run on a proxy instead: open API Settings, choose Proxy, add a configuration with Kyrune Mini's three values and send your message again. Switch back to JanitorLLM whenever you like.
Switching helps only when JanitorLLM itself is the problem. If the proxy then answers with an error, switching again will not fix it. Look the error up in the list below: it may be a wrong URL, key or model id, a chat too long for the model, a reply still running, or a daily limit.
This page does not show live JanitorLLM status. JanitorAI posts its own updates in its changelog.
Kyrune Mini runs on Kyrune's own GPUs. The longest a request waits in line for its first words is 3 minutes; after that it times out.
Tested on kyrune.com with a Free test account
One request to https://kyrune.com/api/v1/chat/completions, sent the way JanitorAI sends it (streamed), with model kyrune-mini-1.0-free.
- Character
- Mira, the cheerful keeper of a tavern in a rainy harbour town.
- Prompt
- *I push open the tavern door and shake the rain off my cloak.* Evening. Got a room for the night?
- Reply, trimmed
- *I wipe down a polished mug with a linen cloth, looking up with a bright smile as the bell above the door jingles.* "Evening! Oh, you've come in from quite the deluge, haven't you?" [...] "Aye, I've a cozy room left with a dry blanket and a view of the rain. Are you looking for just a bed, or did you want supper to warm your bones first?"
- First words after
- 1.2 seconds (whole reply in 3.0 seconds)
If the connection fails
| You see | What it means |
|---|---|
401 No API key | The app sent no key. Paste your API key into the app's key field. |
401 This API key doesn't work | The key is mistyped or revoked. Copy it again from the API page. |
404 Wrong address | The URL is wrong or doubled, for example /v1/v1. Set the API URL to the one on this page. |
400 No model named 'x' | The model id has a typo. Use kyrune-mini-1.0-free. |
413 This chat is too long for the Free plan (8,000 tokens) | On another model, the chat is longer than a free prompt. Start a new chat, or use Kyrune Mini. |
413 This chat is too long for this model | The chat no longer fits the model's context, prompt and reply together. Nothing is cut or summarized for you. Start a new chat or delete older messages. |
429 Your last reply is still being written | On Free, your account answers one reply at a time. Wait for it, or stop it, then send again. |
429 Kyrune Mini daily limit reached (100 requests) | Kyrune Mini's requests for today are used up. They reset at 00:00 UTC. |
403 Daily limit reached (10 requests on Free) | Another model used up the Free plan's 10 a day. Switch to kyrune-mini-1.0-free, which has its own 100 a day, or wait until 00:00 UTC. |
Free and low-cost roleplay models
How this list is made: Kyrune Mini, plus every Roleplay API model with no daily cap from Plus, from the live model list on . On the Free plan all of them except Kyrune Mini count toward the 10 a day. Listed in the API's own order, not ranked by quality. See every model in Models.
| Model | Context | On the Free plan | No daily cap from |
|---|---|---|---|
Kyrune Mini 1.0kyrune-mini-1.0-free | 28,672 | Free, 100 a day of its own | 300 a day on paid plans |
Euryale 70B v2.3euryale-70b-v2.3 | 32,000 | Counts toward 10 a day | Plus, $12.99 a month |
Magnum v4 70B SEmagnum-v4-70b-se | 32,000 | Counts toward 10 a day | Plus, $12.99 a month |
Hanami 70B x1hanami-70b-x1 | 32,000 | Counts toward 10 a day | Plus, $12.99 a month |
Magnum v4 72Bmagnum-v4-72b | 32,000 | Counts toward 10 a day | Plus, $12.99 a month |
Strawberry Lemonade 70B v1.1strawberrylemonade-70b-v1.1 | 32,000 | Counts toward 10 a day | Plus, $12.99 a month |
Anubis 70B v1.1anubis-70b-v1.1 | 32,000 | Counts toward 10 a day | Plus, $12.99 a month |
Kunou 72B v1kunou-72b-v1 | 32,000 | Counts toward 10 a day | Plus, $12.99 a month |
Privacy
We do not store what you send through the API or what comes back. We store that a request happened, to which model, how large it was and how long it took.
We do not train AI models on your prompts.
More in the Privacy Policy.
Common questions
On kyrune.red, for adults 18 and over, yes: Kyrune adds no content filter to Kyrune Mini's replies, and sexual content between adults is allowed. Sexual content involving a minor or anyone who looks like one, sexual content without consent, and real people without their agreement are never allowed.
Yes. Kyrune Mini 1.0 is free on every plan, the Free plan included, with requests of its own that never count toward your plan's daily requests: 100 a day on Free and 300 a day on paid plans. You need a free Kyrune account and an API key.
Kyrune Mini, model id kyrune-mini-1.0-free, through JanitorAI's Proxy setting with the URL https://kyrune.red/api/v1/chat/completions and a free API key. It helps when JanitorLLM itself is failing or slow; it does not fix a wrong URL, key or model id.
On Free, Kyrune Mini takes 100 requests a day, answers one reply at a time and reads up to 28,672 tokens, prompt and reply together. Every other model counts toward 10 requests a day, which reset at 00:00 UTC.
kyrune-mini-1.0-free. Type it in JanitorAI's Model Name field, or pick it from the model list in SillyTavern and Chub.
In JanitorAI, https://kyrune.red/api/v1/chat/completions. In SillyTavern and Chub, https://kyrune.red/api/v1.
Up to 28,672 tokens on every plan, shared by the prompt and the reply. Nothing is summarized or cut to make a chat fit; a chat that no longer fits is refused with a message that says so.
No. Kyrune Mini runs with thinking off.
No. We do not store what you send through the API or what comes back, and we do not train AI models on your prompts.